Google Antigravity CLI replaces Gemini CLI by June 18 — key migration steps

Google has set a firm deadline of June 18, 2026, for developers to transition from the legacy Gemini CLI to its successor, Antigravity CLI. After this date, requests will no longer be processed for users on free tiers or Google AI Pro and Ultra subscriptions. The change, announced at Google I/O 2026, has flown under the radar despite its potential to disrupt workflows that rely on terminal-based AI agents.

Why Google is retiring Gemini CLI

At the heart of this transition is a fundamental shift in how developers use AI agents. Google’s reasoning highlights a mismatch between the original Gemini CLI’s design and modern workflows. The legacy tool was built for single-agent, terminal-bound tasks, but today’s development environments demand multi-agent collaboration, parallel processing, and seamless backend integration. Antigravity CLI was engineered to meet these new requirements from the ground up.

The announcement at Google I/O 2026 introduced Antigravity 2.0 as a cohesive agent-first development platform, encompassing four core components:

• Antigravity 2.0: A rebuilt desktop application
• Antigravity CLI: A Go-based terminal tool replacing Gemini CLI
• Antigravity SDK: For building custom agent workflows
• Managed Agents API: Enables spinning up autonomous agents with a single API call

While much of the coverage focused on the desktop app and the new Gemini 3.5 Flash model, the critical change for most developers lies in the CLI transition.

Key changes in Antigravity CLI

A complete rewrite in Go

The Antigravity CLI binary, now named agy, represents more than a simple rename. It has been entirely rewritten in Go, eliminating runtime dependencies and delivering faster response times. Developers accustomed to running gemini commands in shell scripts, GitHub Actions, or Dockerfiles will need to update every invocation before the deadline.

For most projects, migration involves a straightforward find-and-replace operation. However, Dockerfiles that reference GEMINI.md files require explicit file mapping adjustments to ensure compatibility with the new .agents/ directory structure.

Configuration and directory restructuring

The migration extends beyond binary names to the project’s configuration and skill directories:

• GEMINI.md becomes AGENTS.md
• .gemini/skills/ moves to .agents/skills/

A significant advantage during this transition is the ability to run both gemini and agy simultaneously. Each binary maintains separate configuration trees and authentication flows, allowing developers to test Antigravity CLI on specific tasks while keeping legacy workflows intact until June 18.

Shift to asynchronous multi-agent processing

The most transformative change is Antigravity CLI’s handling of complex tasks. Unlike the legacy tool, which locked terminals during execution, the new CLI uses an asynchronous background processing model. Developers can initiate large-scale refactors or multi-source research tasks without losing control of their terminal sessions.

Antigravity supports up to four or five parallel agents before performance begins to degrade. Multi-agent orchestration is defined in AGENTS.md, a markdown file that outlines agent roles, communication protocols, and task distribution. The syntax will feel familiar to users of Claude Code and its CLAUDE.md configuration pattern.

The power of the SKILL.md system

A standout feature of Antigravity is its Skill system, which deserves closer attention amid broader platform discussions. In this framework, a Skill is a directory-based package containing a SKILL.md file and optional supporting scripts or templates. The design prioritizes simplicity, relying solely on markdown and YAML—formats developers already know.

A typical skill structure looks like this:

.agents/skills/
└── my-skill/
    ├── SKILL.md       # Instructions and rules
    └── scripts/
        └── run.sh     # Optional supporting script

The distinction between Skills and Rules is pivotal. Rules in Antigravity are persistent, functioning like system-wide prompts that apply across all agent interactions. Skills, by contrast, are on-demand and load only when triggered. This design prevents bloating agent interactions with unnecessary capabilities.

For frontend developers, this translates to significant productivity gains. Instead of repeatedly explaining coding conventions to an AI, conventions are defined once in markdown, stored in .agents/skills/, and consistently applied across every project using that skill. The agent no longer needs to relearn your workflow every session.

Who needs to migrate—and how

Several user segments face the June 18 deadline:

• Free-tier and Google AI Pro/Ultra users: Must migrate before the deadline; no exceptions.
• Enterprise users on Gemini Code Assist Standard or Enterprise licenses: Unaffected; migration can proceed at their convenience.
• Enterprise Gemini Code Assist for GitHub: Must complete migration by June 18; no new installations permitted afterward.

Google estimates the migration process takes under ten minutes for most setups. The essential steps are:

1. Install Antigravity CLI using the official instructions.
2. Import existing plugins with agy plugin import gemini.
3. Rename configuration files: mv GEMINI.md AGENTS.md.
4. Move skill directories: mv .gemini/skills/ .agents/skills/.
5. Update CI/CD scripts by replacing all gemini references with agy.

After completing these steps, test a real workflow to confirm compatibility before the deadline.

The clock is ticking, but the path forward is clear. Developers who act now can avoid disruptions and unlock the full potential of Antigravity’s agent-first platform.